Subh775/step_scheduler.rf-detr-nano

RF-DETR with step learning rate scheduling and optimized hyperparameters

We fine-tuned RF-DETR using a step learning rate scheduler on a custom dataset. Within two epochs, the model achieved a +3.7 increase in mAP@50:95, with balanced improvement in classification and localization losses. EMA weights consistently outperformed standard parameters, indicating stable convergence. Per-class analysis shows strong performance on well-represented categories like two-wheelers and trucks, while smaller or visually ambiguous classes such as minibuses remain challenging, suggesting future improvements via data balancing.

Total Loss over epochs

Per-class mAP@50:95

Per-class Precision and Recall (Last Epoch)

COCO mAP vs Epochs